A Fully Integrated, High Speed, Motion Controller Hardware Accelerator for Autonomous Systems

Efficient and accurate motion control for any autonomous system (such as, aerial vehicles, ground robots) is a major challenge for robotics researchers and enthusiasts. Fast moving systems demand accurate control of motion for robust operation. To achieve this, high loop rate with fixed minimum jitter is required. As the size of these robots shrinks, they become more agile and are capable of performing very aggressive maneuvers. For such operations, the overall control loop timing requirements become critical. In state-of-the-art approaches, the problem of motion control is solved using a dedicated microcontroller with a specialized software stack. However, in this approach, the control loop suffers from jitter. Also, this approach fails for a multi-axis system (with increased number of axis in the system). Further, this approach increases the hardware component on the platform which adds to the total cost and weight. These challenging requirements demand a dedicated hardware accelerator for efficient operation. To this end, we devise an architecture and data flow for the design of motion controller hardware accelerator for the autonomous system use cases. This accelerator implements the complete control loop in hardware and allows multiple controllers (hence multiple actuators or motors) to run in parallel. Motion controller (MC) hardware accelerator was implemented and integrated on a test System-on-Chip (SoC) and the performance was fully characterized. Further, we have demonstrated the hardware accelerator operation on a physical mobile autonomous robotic system (ground and aerial vehicle). The experimental results demonstrate efficient control operation and our solution achieves ∼44x faster response time and ∼1000x improvement in overall power consumption compared to the software based solution.